1. Field of the Invention
The present invention relates to a fuel injection controller that performs injection amount learning of an engine.
2. Description of Related Art
Conventionally, a method of performing a pilot injection of injecting an extremely small amount of fuel before a main injection is known as a measure for reducing a combustion noise or for inhibiting NOx of a diesel engine. In the case of the pilot injection with a small command value of the injection amount, improvement of accuracy of the minute amount injection is required in order to fully exert the effects (i.e., reduction of combustion noise and inhibition of NOx). Therefore, injection amount learning of sensing a deviation between the command injection amount of the pilot injection and the actually injected fuel amount (i.e., actual injection amount) and performing correction on a software side is needed.
Therefore, a following technology is proposed, for example, in JP-A-2005-36788. That is, in a M/T vehicle, a single-shot injection (i.e., injection for learning) is performed at a time of no-load such as a time of shift change (i.e., at time of clutch-cut) and of no-injection during deceleration. The actual injection amount at that time is estimated from a change of an engine rotation state (for example, amount of fluctuation of engine rotation speed). An injection amount characteristic deviation of an injector is calculated based on the thus-estimated actual injection amount and the command injection amount. Then, the command injection amount is corrected to decrease the injection amount characteristic deviation.
A fuel injection controller described in JP-A-2005-36788 performs the injection for the learning only when the clutch is disengaged, in order to accurately sense the change of the engine rotation state caused by the injection. In an A/T vehicle, such the state (i.e., neutral state) is rare. Therefore, when this scheme is applied to a diesel engine of the A/T vehicle, there are few opportunities of the injection amount learning.
So, in the case of the A/T vehicle, it is proposed to perform the injection amount learning when an engine side and a transmission side (extending from transmission to wheels) are engaged but are not locked up. However, disturbances such as a road noise and torsional vibration are added to the engine side from the transmission side. Therefore, the sensing accuracy of the change of the engine rotation state worsens, so the learning accuracy worsens.
Even in the case where the scheme is applied to the diesel engine of the M/T vehicle, if a damper is attached to a flywheel, a delay or the like can arise in the fluctuation of the engine rotation speed caused by the injection for the learning. Accordingly, the sensing accuracy of the change of the engine rotation state worsens, so the learning accuracy worsens.